This study describes a method of modeling human trunk and whole body backward bending and suggests a possible neural control strategy. The hypothesis was that the control system can be modeled as a linear feedback system, in which the torque acting at a given joint is a function of the state variables (angular positions and angular velocities). The linear system enabled representation of the feedback system by a gain matrix. The matrix was computed from the kinematics recorded by a movement analysis system and from the joint torques calculated by inverse dynamics. To validate the control model, a comparison was made between the angular kinematics yielded by the model and the experimental data. Moreover, for all subjects, the same relationships between feedback coefficients were found although gain values were different. The study showed that the feedback system is an appropriate model of the strategy from performing an accurate controlled trunk or whole body backward bending in the sagittal plane.